Duet Song and Sex Roles During Territory Defence in a Tropical Bird, the Black-Bellied Wren, Thryothorus Fasciatoventris

ANIMAL BEHAVIOUR, 2004, 68, 721–731 doi:10.1016/j.anbehav.2003.10.026

Duet song and sex roles during territory defence in a tropical bird, the black-bellied wren, Thryothorus fasciatoventris

DAVID M. LOGUE & DAVID E. GAMMON Department of Biology, Colorado State University and Smithsonian Tropical Research Institute, Republic of Panama

(Received 18 February 2003; initial acceptance 11 August 2003; ﬁnal acceptance 21 October 2003; published online 23 August 2004; MS. number: A9557)

A diverse array of bird species show a behaviour known as duet singing, in which mated pairs sing temporally coordinated songs. Several studies have shown that simulating territory intrusion with conspecific song playback evokes duet song from duetting bird species, but the adaptive significance of coordinated song during territorial defence remains unclear. The function of duetting is further obscured by our poor understanding of the roles of the sexes in territory defence among year-round territorial species. We developed a simple optimality model that predicts the conditions under which intersexual territorial defence is most likely to occur. We used male solo song and female solo song playback as well as a novel method called ‘stereo duet playback’ to test the predictions of the model and to explore the function of song initiation and song answering during territorial encounters. We conducted these experiments in the field on the black-bellied wren, a year-round territorial passerine. Birds of both sexes responded to all treatment types. Males initiated more songs during opposite-sex playback than during same-sex playback and both sexes were more likely to answer their mates’ songs when mates were physically closer. We argue that the acoustic mate-guarding hypothesis does not adequately account for these results, and suggest that duetting during territorial encounters allows mates to identify one another, thus preventing intrapair aggression. As predicted by our model, females showed a strong same-sex bias in territory defence, whereas males approached playback of both sexes. Also in support of the model, males with dependent juveniles showed stronger intersexual territoriality than males not involved in a breeding effort. Ó 2004 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Avian vocal duetting is especially common in species in versus heterospecific playback (Levin 1996b). Magpie lark, which a mated male and female defend a shared territory Grallina cyanoleuca, pairs coordinate a higher proportion of throughout the year (Farabaugh 1982). Although the their songs to form duets during conspecific playback adaptive significance of this behaviour varies among versus prior to playback (Hall 2000). Similar results have species, sexes (Levin 1996b) and contexts (Sonnenschein been found for stripe-backed wrens, Campylorhynchus & Reyer 1983), simulated territory intrusion evokes height- nuchalis (Wiley & Wiley 1977), usambiro barbets, Trachy- ened levels of duet song in many species of year-round phonus usambiro, slate-coloured boubous, Laniarus funebris, territorial birds. Pairs of robin chats, Cossypha heuglini, duet and striped kingfishers, Halcyon chelicuti (Wickler 1976). In when presented with conspecific mounts and playback all of these experiments, an elevated level of duet song (Hultsch & Todt 1984). Female bay wrens, Thryothorus during playback was associated with other measures of nigricapillus, initiate song more often and are more likely to agonistic response (e.g. flights, approach to speaker). form duets by answering their mates during conspecific It is clear that pairs duet more during simulated territorial intrusions, but what is the adaptive significance of coordinated song to individual birds? One obstacle to Correspondence: D. M. Logue, Department of Biology, Colorado State answering this question is our poor understanding of University, Fort Collins, CO 80523-1878, U.S.A. (email: dlogue@ territory defence in year-round territorial birds (see Table 5 lamar.colostate.edu). in Hall 2000 for a review of duetting and joint territorial 721 0003–3472/03/$30.00/0 Ó 2004 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. 722 ANIMAL BEHAVIOUR, 68, 4

defence). Several studies have reported a strong same-sex intersexual territoriality when retention of the current bias in territorial defence (Greenberg & Gradwohl 1983; mate stands to provide high fitness benefits. This is likely Freed 1987; Levin 1996b; Morton & Derrickson 1996; Hall to be the case during the period of nestling and juvenile 2000; Bard et al. 2002; Seddon et al. 2002). The evidence care if stepparents do not aid in raising the young (as in presented in such studies typically shows that birds female black-bellied wrens; D. M. Logue, unpublished respond more strongly to intrasexual versus intersexual data) or if they commit infanticide (as in male house solo song playback. All of these studies, however, also wrens, Troglodytes aedon; Freed 1987). revealed that mated birds respond physically (e.g. by The model does not address mate following during approaching the speaker) to opposite-sex playbacks. It opposite-sex playback as a means of mate retention, has been suggested that this apparent intersexual territory although in some cases it may provide a simpler explana- defence represents intersexual protection against infanti- tion for the observed pattern of intersexual approach to cide (Freed 1987) or a means of mate retention when the playback speaker. In the interest of keeping the model unattended individuals are likely to advertise for a new simple, we did not consider the cost represented by an mate (Morton & Derrickson 1996). injury to the mate who has defended alone. Including We developed a simple model that predicts the con- such a cost would increase the conditions under which ditions under which intersexual territory defence should intersexual defence is less costly than the alternative. The occur. The model assumes strict social monogamy; either model applies equally well to a pair of intruders, rather an intruder is rebuffed or it replaces the same-sex territory than a single intruder, if we bear in mind that the cost C to holder. Consider the case in which a solo female intrudes the focal individual is likely to be greater if the individual upon a mated pair’s territory. The paired male will have also must defend against a same-sex intruder. Until now, it a lifetime fitness of V1 if he keeps his current mate and has been impossible to address the question of intersexual a fitness of V2 if the intruder becomes his mate. If the male territoriality during simulated pair intrusions because does not attempt to drive the intruding female from his traditional duet playbacks broadcast both male and female territory, there is a probability p that she will achieve duet contributions from a single speaker. We developed residence on the territory and become his mate. We a method of broadcasting duet recordings from two assume that if he does help to defend the territory against sources to more realistically simulate a duetting pair. the intruder, he will retain his current mate. Finally, We conducted song playback experiments on free-living defending the territory against the intruding female black-bellied wrens to address the following questions. (1) imposes a cost C to the male. With these assumptions, How and why do these birds initiate song and answer we can generate a simple model that predicts that a mated their partners’ songs to form duets during territorial male will optimize his fitness by defending his territory intrusions? (2) Does either sex show intersexual territorial against intruding females when defence? Black-bellied wrens maintain all-purpose territories throughout the year. Both males and females sing repertoires of sex-specific song types, some of which they V1 C > pV2Cð1 pÞV1 share with other individuals in the population. These The left side of the inequality represents the fitness payoff songs may be sung independently or coordinated with the to the male if he defends the territory against the intrud- song of the mate to form a duet. In contrast to the ing female. The right side of the inequality represents his congeneric bay wren (Levin 1996b), both sexes of black- average fitness payoff if he does not defend; the average bellied wrens initiate duets. Individuals typically alternate duet contributions to form duet trains (sensu Brown & payoff if the male does not defend will depend on both V1 Lemon 1979) that vary in length from two to 16 and V2 because either female could potentially occupy the territory after the conflict. The inequality can be simpli- contributions. fied to yield Male black-bellied wrens are substantially larger than females (all Ns > 21; Xmass: 28.0 g versus 22.3 g; Xtarsus: 26.7 mm versus 24.0 mm; Xwing: 67.7 mm versus V1 V2 > C=p 63.5 mm, respectively). Unmated males regularly obtain and hold solo territories. We have no evidence that This model suggests that a male should defend his females naturally acquire or maintain solo territories, territory against an intruding female given the conditions nor do we have evidence that older offspring help in the that (1) the male assesses that his fitness with the current rearing of younger offspring. Social monogamy is the only mate would exceed his fitness with the intruder and (2) known mating system in this species. Both sexes are the cost of intersexual territory defence is sufficiently low. highly active in nest building and nestling and juvenile Of course, one could apply the model to a mated female care (personal observation). during a solo male intrusion to make predictions about intersexual territorial defence by females. For species in which one sex is larger than the other, we METHODS expect that the larger sex will pay lower costs for intersexual territoriality than the smaller sex. Therefore, all Study Population else being equal, the model predicts that in species where one sex is larger, the larger sex is more likely to practice We performed a series of field playback experiments on intersexual territoriality. The model also predicts increased 16 pairs of free-living black-bellied wrens near Gamboa, LOGUE & GAMMON: VOCAL DUETTING AND TERRITORIAL DEFENCE 723

Republic of Panama. Experiments were conducted from 15 parabolic reflector. During the 2002 season, the recordings May to 15 August 2002, a period that corresponded with used as playback stimuli were made with a Sony Minidisc the tropical wet season and the approximate middle of the recorder (MZ-R700) and a Sennheiser omnidirectional black-bellied wren’s breeding season. Four of the pairs microphone (ME62) mounted in a 60-cm Telinga Pro- lived in the Gamboa Woods, a small (w25 ha) wood lot universal parabola. True pitch for the recordings was surrounded by residential areas on three sides. The verified using a tuning fork. remaining 12 pairs lived near railroad tracks that run Using the program Real Time Spectrogram (Kay Eleme- parallel to the Panama Canal, on its south side. Creation trics, version 2.3), we digitized the recorded songs at of the railroad generated swaths of second-growth and a sampling rate of 44 100 Hz and visually examined edge forest on both sides of the tracks. Black-bellied wrens sonagrams for quality. When choosing recordings, we are partial to such regenerating and edge habitats (Ridgely attempted to maximize signal-to-noise ratios and mini- & Gwynne 1989). Both study sites were characterized by mize reverberation. We used cursor-defined filters in mixed-level forests with dense undergrowth and high Syrinx (John Burt: http://syrinxpc.com/) and Canary densities of tangled woody vines, a preferred foraging (Charif et al. 1995) to remove low-frequency noise. Filters substrate. were never used to remove high-frequency noise because Prior to the study, permits for conducting song playback we did not want to delete the high-frequency harmonics, experiments and for banding birds were obtained from which have been shown to contribute to sound-source the Colorado State University Animal Care and Use localization in this genus (Naguib 1995). After filtration, Committee and from Autoridad Nacional del Ambiente we adjusted the amplitude of the signals so that the in Panama. Of the 32 birds used in this study, 12 males and maximum amplitude of each signal was the same. eight females were uniquely colour-banded prior to the Because our access to computers changed throughout onset of experimental manipulation. We banded the the course of this study, we used both Canary (which runs remaining four males and five of the remaining eight on a Macintosh platform) and Syrinx (PC) to create stereo females before the end of the study. Over two field seasons duet stimuli. We believe that stimuli made with the two of field observations, we never observed pairs to move programs are indistinguishable. In this report, we describe together to a new territory. We have observed five instances how to create stimuli using Canary, although the process in which females left (or died) and were quickly replaced, is similar to the one that we employed using Syrinx. but these were followed by greatly heightened levels To overcome the limitations inherent in the use of of spontaneous duet song. Despite frequent observation a single sound source to simulate a paired intrusion (see periods (at least twice weekly), we did not encounter any Introduction), we converted all duet recordings into stereo drastic changes in the level of spontaneous duet song, so it stimuli (Fig. 1). This allowed us to broadcast the male and was unlikely that we misidentified individual birds. female contributions through separate speakers. Both We observed dependent juveniles on six of the 16 male-initiated duets (MIDs) and female-initiated duets territories during the 2002 season. The presence of (FIDs) were converted into stereo stimuli. All duet stimuli juveniles was unambiguous because males produced loud consisted of three contributions with mates alternating ‘cream-o-wheat’ (D. M. Logue, unpublished observation; contributions (MFM or FMF). We used Canary to display onomatopoetic device after F. O. Chapelle cited in Ridgely two identical sonagrams of the high-pass-filtered record- & Gwynne 1989) calls almost continuously while attend- ings. Using the cursor-delimited filter, we erased all of the ing to dependent juveniles but used this call rarely in the male contributions from one of the copies and all of the absence of juveniles. Also, the presence of juveniles is female contributions from the other copy. Subsequently, readily apparent from their distinctive ‘pizeet’ calls and we set the sonagrams’ clipping levels to the lowest settings their grey (rather than white) throat feathers (Ridgely & that maintained a legible signal. By examining the Gwynne 1989; Stiles & Skutch 1989). During the study, we adjusted sonagram and listening to the signal, we de- sexed the birds by morphology and song. Subsequent termined that our filtering had eliminated the unwanted amplification and visualization of the sex-chromosome- signals. We then copied the two files and pasted them into linked CHD gene with 1237L and 1272H primers (Kahn the two channels of a blank stereo recording, maintaining et al. 1998) from a subset of 10 of the birds (seven males the original temporal relationships among the three and three females) confirmed that those individuals were contributions. For each duet stimulus, a coin flip de- correctly sexed. termined which channel would carry the female and male contribution(s). We are confident that we correctly distinguished male Stimuli contributions from female contributions when creating stereo stimuli. Male and female contributions differ in All of the recorded vocalizations used as playbacks were timbre and pattern of frequency modulation, so the obtained during May–August of 2001 and 2002 from free- experienced listener can accurately assign syllables to living birds near Gamboa, Republic of Panama. Songs used one or the other sex. Furthermore, duet contributions as stimuli were recorded during or shortly after stimulat- are identical to solo songs and males typically repeat ing the birds with song playback. During the 2001 season, a song type alone several times before incorporating that recordings were made with a Marantz tape recorder (PMD song into a duet. Because a given male song type is highly 222) and a Radio Shack omnidirectional microphone stereotyped, we could compare the solo songs to the duet (product number 33-3014) mounted in a 45-cm aluminium songs to identify ambiguous syllables. 724 ANIMAL BEHAVIOUR, 68, 4

(a)

0 Frequency (kHz)

(b) (c)

0 12 12 Time (s) Figure 1. An example of duet splitting using cursor-delimited filters. (a) A high-pass filtered three-contribution duet is copied twice. This example shows a male-initiated duet. (b) Cursor-delimited filters are used to remove the female contribution from one copy, leaving only the male duet contributions. (c) The male contributions are then removed from the other copy, leaving only the female contribution. The split songs can now be pasted into the two channels of a stereo track and played through separate speakers without altering their temporal relationship (see text).

All solo songs used in playbacks were high-pass filtered inside the territory and 20–30 m away from the bird. We in the same manner as the duets. After filtration, duet and chose areas with sufficient visibility to allow for distance solo stimuli were standardized for amplitude and recorded estimates and with sufficient vegetation to provide on Minidisc. Each Minidisc track contained a single numerous perch sites at varied distances from the speak- stimulus followed by enough blank space to bring the er(s). In the event that we did not detect one of the birds track to 30 s. We played stereo stimuli through a pair of (usually the female) during the playback, we repeated the identical 5-W speakers (Saul Mineroff Electronics, SME- trial on another day. If we observed only one bird on the AFS). The speakers were placed face-up on the ground, 8 m second trial, only the first trial was counted (N Z 2). apart. They were connected to a single Minidisc player via Each trial was preceded by 5 min of observation without a cord composed of a single male stereo plug leading into stimulus (hereafter the ‘pretrial period’). During this two male mono plugs. Solo song playbacks were set up period, we attempted to record the same response varia- similarly, but required only one speaker and a standard bles that we recorded during the playback (see below). connecting cord. Each observer recorded the focal birds’ vocalizations as well as his observations throughout the pretrial and trial periods. One of us (D.E.G.) made recordings with a Senn- Playback Trials heiser omnidirectional microphone (ME62) mounted in a Telinga Pro-universal parabola and connected to a Sony We ran stereo duet playback trials from 29 May to 19 Minidisc recorder (MZ-R700). The other observer (D.M.L.) July 2002, and solo playback trials from 29 June to 7 made recordings using a Sennheiser short shotgun micro- August 2002. We lack complete data on the black-bellied phone (ME-66) and a Marantz cassette recorder (PMD wren’s breeding season, but we observed nests and young 222). If, during the pretrial observation period, we de- juveniles on territories throughout the course of this termined that one or both of the birds were involved in investigation. All trials began between 0600 and 0830 territorial countersinging or nest building, we postponed hours. We avoided playing birds their own vocalizations the trial until the behaviour ceased. At the end of the or the vocalizations of a neighbour, and no stimuli were 5-min pretrial period, one of us began the playback used in more than one trial. Before each experiment (duet stimulus. We matched playback intensity to the intensity playback or solo playback), we randomly determined of the vocalizations given by focal birds. The Minidisc was which stimulus type a pair would receive first. set on ‘one-track repeat’ mode so that the stimulus was A pilot study had shown that black-bellied wrens do not repeated every 30 s for 5 min. We recorded the following consistently respond to playback when the broadcast responses for each of the two focal birds: vocalizations, speaker is placed haphazardly within the territory with distance(s) from the male and/or female speaker, and respect to the focal birds, so we used the following system distance to the mate. Distance measures were estimated for placing the speakers. We approached a territory on using the 8-m speaker cable as a reference. We updated which we wanted to run a trial and listened for black- distances after each change in location greater than 1 m. bellied wren vocalizations. When a bird (usually the male) We each made observations on the bird that was closest was located, we set up the playback speakers at least 15 m to each of us at the time. Care was taken to communicate LOGUE & GAMMON: VOCAL DUETTING AND TERRITORIAL DEFENCE 725 the location and sex of the animal under observation playback trials. One-tailed statistical tests were appropriate (when known) to the other observer. Occasionally, one of in this instance because all previous song playback studies us was able to see both of the birds, and the other could have documented a stronger approach to same-sex play- not see either bird. When this happened, the observer back than to opposite-sex playback (see Introduction). with the better vantage point recorded observations for Approach distances to the two speakers during stereo duet both birds. playbacks were not independent of one another. The paired t test overcame this problem by subtracting the distance to the opposite-sex speaker from the distance to Data Compilation the same-sex speaker, generating a single variable that described the magnitude and direction of approach bias. Each trial generated two recordings, which were used to After careful consideration, we chose to conduct uni- fill in a timeline with the events that transpired. We first variate, rather than multivariate, comparisons of behav- analysed the Minidisc recording because this format iour among treatment types. Some of our measures were includes a run-time display that can be used to assign correlated with one another, but this does not detract a time to each event. The data from the tape recording from the suitability of univariate analysis as long as the were then added to the timeline. When our distance different measures are understood to represent ‘different estimates differed, the data from the observer with the aspects of the response elicited by playback’ (McGregor better vantage point were used. In the event that it was 1992). not clear which observer had a better vantage point, the Closest approach, number of songs initiated and answer distances were averaged. rate were used as dependent variables for ANOVA models. We used the timeline to score the following response We constructed ANOVA models using stimulus type, variables for each bird: song initiations, answers to the presence of dependent offspring and an interaction term mate’s songs, closest approach to each speaker, and the as our treatment factors. If the interaction term was not distance between the mated birds (hereafter referred to as found to have at least a marginally significant effect the interbird distance or IBD). We were only rarely able to (P ! 0.10), it was dropped from the model. For the sake measure the IBD during the pretrial period; therefore, we of clarity, we report these statistics only in those cases do not use these data in comparisons of pretrial versus where the presence of offspring was found to have trial behaviour. Vocal initiations included all tonal vocal- a significant effect. If the presence of juveniles was not izations that did not immediately follow a vocalization by found to have a significant effect on the results, data were the focal bird’s mate. An answer was any vocalization that pooled with respect to presence of offspring. In these immediately followed a mate’s vocalization, thus forming cases, we used two-way ANOVAs to test for treatment a duet. Only one answer was scored for each duet. effects while blocking for individual identity. We were interested in the frequencies with which the We used multiple regression to explore answer rates in sexes answered their mates to form duets. We derived relation to IBD and average distances to the speakers. We answer rates by dividing the number of times an in- did not use a model-selection algorithm (e.g. stepwise dividual answered its mate by the number of initiations by model selection) because the various spatial variables the mate (after Levin 1996b). Derivation of answer rate were nonindependent and it was our aim to determine required that the focal bird’s mate initiated a song to the usefulness of each spatial variable in predicting which it could respond. Failing this requirement, we answer rates while holding all other spatial variables logged a missing value. constant. Because we conducted all of the duet playback experi- Statistical Analysis ments before beginning solo playback experiments, it was not possible to control for the effects of order of pre- Most response variables required transformation prior sentation when comparing the birds’ responses to duet to analysis with parametric statistics. Arcsine transforma- versus solo playbacks. Accordingly, we did not analyse the tions were used for percentage data and ln(X C 1) trans- data in this manner. During the female-initiated duet trial formations were used for variables with large extreme on the mated pair F–BG, the male (F) remained more than values. All statistical analyses were conducted with Mini- 40 m away from the speaker, and his pattern of song tab (2000) software. We adhered to the standard alpha suggested that he did not hear the playback. Data from level of 0.05, but because sample sizes were small, we this trial were omitted from all analyses. report P values where 0.05 ! P ! 0.10 as ‘marginally significant’. We compared male and female behaviour prior to RESULTS playbacks using ANOVA models blocking for the identity of the pair. Two-tailed paired t tests were used to compare Pooling all pretrial observations, we found that males closest approach (as measured from the nearer speaker for initiated song at a much higher rate than females (XGSE: stereo duet playbacks), initiation and answer rate during males: 2.46 G 0.32 song initiations/min; females: pretrial and trial periods. 0.06 G 0.04 song initiations/min; ANOVA blocking for Z ! We used one-tailed paired t tests to test the hypothesis pair: F1,93 150, P 0.001). Males were marginally more that birds would approach the same-sex speaker more likely to answer their mates to form duets, answering closely than the opposite-sex speaker during stereo duet 47.8% of their mates’ songs, compared with an answer 726 ANIMAL BEHAVIOUR, 68, 4

Table 1. Summary of results of male-initiated duet (MID) and female-initiated duet (FID) playbacks 30 (a)

Comparison Male subjects Female subjects 20

Trial vs pretrial 10 Distance to speaker Decreased Decreased (4) Initiation rate Unchanged Increased 0 (always high) (–8) Answer rate Not testable Increased* –10 MID vs FID Distance to speaker No effect No effect –20 Initiation rate No effect No effect Answer rate No effect No effect –30 Pooled over treatments Speaker bias No bias Female speaker 30 bias (b) Answer rate Negative Negative 20 vs IBD correlation correlation

‘IBD’ represents the distance separating a mated pair. 10

*Data for male- and female-initiated duet playback treatments were Trial initiations – pretrial pooled. 0 (0) (1)

–10 rate of 16.3% for females (ANOVA blocking for pair: Z Z F1,37 3.81, P 0.059). –20

Comparisons to Pretrial Behaviour –30 FS playback MS playback Birds of both sexes responded strongly to both duet and Figure 2. Boxplots showing variation in the amount of vocal solo playbacks (summarized in Tables 1, 2). Males ap- initiations during both types of solo playback for (a) male and (b) proached the speaker more closely during all trial types female subjects (FS Z female song, MS Z male song). We sub- than during control periods ( paired t tests: all P % 0.01). tracted the number of initiations during the pretrial period from the Compared with pretrial (Pre) levels, male song initiation number of initiations during the trial period to obtain input values for rates were marginally lower during male song (MS) the boxplot. Medians (horizontal line within the box and number in playback (XGSE: Pre: 14.4 G 3.0 songs; MS: 5.3 G 1.3 parentheses), quartiles (top and bottom of box), the 0.05 and 0.95 quantiles (tips of vertical whiskers), and extreme data points songs; paired t test: t12 Z 2.07, P Z 0.063; Fig. 2a) and marginally higher during female song (FS) playback (asterisks) are shown for each boxplot. (XGSE: Pre: 13.6 G 4.0 songs; FS: 22.1 G 4.4 songs; paired t test: t12 Z 1.882, P Z 0.087; Fig. 2a) but were not affected by either duet playback. We were not able to MID: t5 Z 1.93, P Z 0.1). Our failure to detect female compare male answer rate during the pretrial periods with approach during MID trials may represent type II error, those during the trials because females rarely initiated because missing values reduced our usable sample size to song prior to playback. six, and those samples showed a nonsigniﬁcant tendency We detected signiﬁcant reductions in female distance to towards reduced distance. Female song initiation rates the speaker during both FS and FID trials ( paired t tests: increased during MS, MID and FID trials ( paired t tests: Z Z Z Z Z Z Z Z FS: t9 4.47, P 0.002; FID: t8 3.68, P 0.006), MS: t11 3.88, P 0.003; MID: t15 7.60, P 0.001; Z Z Z Z but not in MS or MID trials (MS: t8 1.71, P 0.13; FID: t14 5.21, P 0.001; Fig. 2b), but did not change

Table 2. Summary of results of male song (MS) and female song (FS) playback

Comparison Male subjects Female subjects

Trial vs pretrial MS FS MS FS Distance to speaker Decreased Decreased Unchanged Decreased Initiation rate Decreased* Increased* Increased Unchanged Answer rate Not testable Not testable Unchanged Increased MS vs FS Distance to speaker With juveniles: closer approach during FS Closer during FS Without juveniles: no effect Initiation rate Higher during FS No effect Answer rate No effect Higher during FS

*Marginally signiﬁcant result. LOGUE & GAMMON: VOCAL DUETTING AND TERRITORIAL DEFENCE 727

in response to FS playback ( paired t test: t11 Z 1.67, playback treatment type. Taken together, these measures P Z 0.12; Fig. 2b). Females answered the songs of their suggest that male response to duet playback is indepen- mates at a significantly higher rate during FS and FID trials dent of the sex that initiates the stimulus duet. ( paired t tests: FS: t9 Z 2.34, P Z 0.04; FID: t10 Z 4.87, P Z 0.001), but not during MS or MID trials ( paired t tests: Female response Z Z Z Z MS: t9 0.69, P 0.50; MID: t9 0.35, P 0.74). In During playback of MIDs, females did not approach the summary, both males and females responded to all female speaker significantly more closely than the male treatment types with a significant change in one or more speaker (XGSE: male speaker: 7.2 G 1.1 m; female speaker: G Z Z of the three measured variables. 6.7 1.1 m, one-tailed paired t test: t17 0.57, P 0.29). During FID playback, they approached the female speaker Response to Stereo Duet Playback more closely than the male speaker (XGSE: male speaker: 8.5 G 0.77 m; female speaker: 7.0 G 1.2 m, one-tailed Male response paired t test: t15 Z 2.50, P Z 0.013). The level of bias did During both MID and FID trials, males did not approach not differ significantly between treatment types (two-way Z Z the male speaker more closely than the female speaker ANOVA blocking for pair: F1,14 0.60, P 0.45), so we (closest approach by males over all duet playbacks: XGSE: pooled over both playback types and found that females male speaker: 7.0 G 0.63 m; female speaker: 6.5 G 0.61 m; approached the female speaker more closely than the male Z Z one-tailed paired t tests: MID: t16 0.02, P 0.49; FID: speaker (one-tailed paired t test: t31 Z 1.85, P Z 0.037), Z Z t15 0.78, P 0.78; Fig. 3a). We detected no difference in especially during trials when the female approached both the males’ closest overall approach to a speaker during speakers closely (Fig. 3b). Duet type did not affect the MID versus FID playbacks (two-way ANOVA blocking for females’ approach to the speaker (two-way ANOVA block- Z Z Z Z pair: F1,14 1.78, P 0.20). Male initiation rates (two- ing for pair: F1,14 0.85, P 0.37), song initiation rate Z Z way ANOVA blocking for pair: F1,14 2.37, P 0.15) and (two-way ANOVA blocking for pair: F1,14 Z 0.46, answer rates (two-way ANOVA blocking for pair: P Z 0.51), or answer rate (two-way ANOVA blocking for Z Z Z Z F1,11 1.31, P 0.28) were also unaffected by duet pair: F1,10 0.39, P 0.54). Like males, females responded to stereo duet playbacks without regard to the sex that initiated the stimulus. (a) 6 Answer rate and IBD We discovered strong negative relationships between the distance separating pair members and the rate at 4 which mates’ songs were answered for both sexes, even when simultaneously accounting for both average and closest distance to both of the playback speakers (multiple regression: male: bYj Z 0.54, tIBD Z 3.78, P Z 0.001; fe- 2 Z Z Z male: bYj 0.48, tIBD 3.50, P 0.002; Fig. 4). Table 1 summarizes the results of the stereo duet playbacks.

0 –8 –4048 Response to Solo Song Playback

Frequency 9 (b) Male response The effect of treatment type (MS versus FS) on males’ approach to the speaker depended on the presence of G 6 dependent juveniles on their territories (X SE: MS with juveniles: 12.7 G 3.8 m; MS no juveniles: 8.2 G 0.94 m; FS with juveniles: 4.3 G 0.78 m; FS no juveniles: 9.9 G 2.4 m; two-way factorial ANOVA: playback type: Z Z Z Z 3 F1,20 5.07, P 0.036; juveniles: F1,20 0.58, P 0.45; playback type)juveniles: F1,20 Z 1.18, P Z 0.016). Specif- ically, males without juveniles approached equally close during both treatments, whereas those with juveniles 0 –8 –4 0 48 approached closer during FS versus MS (Fig. 5a). Males Bias towards female speaker initiated song at higher levels during FS versus MS playback (XGSE: MS: 5.3 G 1.3 songs; FS: 22.1 G 4.4 Figure 3. Histograms showing same-sex approach bias in (a) focal Z songs; two-way ANOVA blocking for pair: F1,11 7.96, males and (b) focal females during both types of stereo duet Z playbacks. The X axis represents the closest approach of an individual P 0.017; see also Fig. 2a). Male answer rate was not G G to the male speaker (in metres) minus the closest approach of the affected by treatment (X SE: MS: 0.80 0.12; FS: same individual to the female speaker. The four trials in which the 0.61 G 0.18; two-way ANOVA blocking for pair: focal female did not approach within 10 m of either speaker are F1,5 Z 0.80, P Z 0.41; Fig. 6a), but because of several miss- indicated in black. ing values we do not have strong conﬁdence in this result. 728 ANIMAL BEHAVIOUR, 68, 4

1 (a) 15 (a) R2 = 31.8% 0.8 MS playback

0.6 10

0.4

0.2 FS playback 5

0 0 5 10 15 15 (b) MS playback 1 Answer rate (b) R2 = 30.9% 0.8 Closest approach to speaker (m) 0.6 10

0.4 FS playback 0.2 5

0 No juveniles Juveniles Figure 5. Interaction plot showing the closest approach to the 0 5 10 15 speaker for (a) male and (b) female subjects with and without Average IBD (m) juveniles during playbacks of male song (MS) and female song (FS). Figure 4. Relationship of (a) male and (b) female answer rates with distance between mates (IBD Z interbird distance). Although the ﬁgure shows a simple regression, the relationship was still signiﬁcant (i.e. MIDs) during FS versus MS playbacks (XGSE: MS: when distances between the birds and the speaker were included in 1.5 G 0.54 MIDs; FS: 8.5 G 1.83 MIDs). See Table 2 for the regression equations (see text). Data are from both male- a summary of results from the solo song playbacks. initiated duet and female-initiated duet trials, so each bird is represented by two points.

DISCUSSION

Female response Our study addressed the following two questions. (1) How Unlike that of males, the closest approach of females and why do male and female black-bellied wrens initiate was not affected by the presence of juveniles (XGSE: MS song and answer their partners’ songs during territorial with juveniles: 14.3 G 3.2 m; MS no juveniles: 13.9 G intrusions? (2) Does either sex show intersexual territory 1.9 m; FS with juveniles: 6.8 G 2.3 m; FS no juveniles: defence? 7.9 G 2.1 m; two-way factorial ANOVA: playback type: Males generally initiated songs at a high rate regardless Z Z Z Z F1,20 9.63, P 0.006; juveniles: F1,20 0.02, P 0.88; of whether or not a stimulus was playing. Speciﬁcally, playback type)juveniles: F1,20 Z 0.17, P Z 0.68; Fig. 5b). male song initiation rate was indistinguishable from pre- Regardless of the presence of juveniles, females ap- trial levels during duet and female solo playback and proached the speaker more closely during FS playbacks marginally lower than pretrial levels during male solo than during MS playbacks (XGSE: MS: 14.0 G 1.5 m; FS: playback (Fig. 2a). Males also initiated song at a higher 7.5 G 1.5 m; two-way ANOVA blocking for pair: level during opposite-sex versus same-sex playback. Fe- Z Z F1,11 17.29, P 0.002; Fig. 5b). Female song initiation male song initiation rate was indistinguishable from pre- rates were unaffected by treatment (XGSE: MS: 2.4 G 1.0 trial levels during female solo playbacks, but increased songs; FS: 2.2 G 0.86 songs; two-way ANOVA blocking for during duet and male solo playbacks (Fig. 2b). pair: F1,11 Z 0.001, P Z 0.978; see also Fig. 2b). Females We did not detect a difference in male answer rate across answered their mates at a marginally higher rate during FS playback types or between the playback and pretrial trials (XGSE: MS: 0.27 G 0.09; FS: 0.49 G 0.08; two-way periods, but recall that the usable sample was quite small, ANOVA blocking for pair: F1,10 Z 4.54, P Z 0.059; and male answer rates were consistently high (X > 60% Fig. 6b). Because of the high male initiation rate during for all playback types). Females generally answered at FS trials, the doubling of the female answer rate resulted in a lower rate than males, but their answer rates increased a dramatic increase in the number of answered male songs during duet trials and increased marginally during female LOGUE & GAMMON: VOCAL DUETTING AND TERRITORIAL DEFENCE 729

songs to coordinate their foraging and other nonagonistic 1 (a) (1) activities. The identity hypothesis extends this hypothesis (0.86) by stating that individuals use duet songs to identify and localize their mates during both agonistic and nonagonistic activities. The identity hypothesis builds on well-established work showing that birds can recognize 0.5 individuals on the basis of song alone (reviewed in Stoddard 1996) and can localize singing conspeciﬁcs (e.g. Morton 1982; Naguib 1995). Suppose that a territorial male sees a conspeciﬁc but does not know whether that individual is an intruder or 0 his own mate. If the male sings, and the territorial female answers in a way that communicates her identity and location, she can prevent the male from directing his 1 (b) aggression at her by mistake. Reversing the sexes yields an

Answer rate equally plausible scenario. Several predictions of this hypothesis were supported by the data gathered in the present study and by previous duetting studies. Mate identification may explain why many species duet 0.5 at high levels during song playback (see Introduction). According to this interpretation, birds initiate song to (0.39) stimulate their mates to answer and they answer to prevent aggression from the mate. This framework gen- (0.17) erates several predictions. Initiation rates should be high (but not necessarily exceeding pretrial levels) for both 0 sexes during duet playbacks regardless of whether play- FS playback MS playback back is an MID or FID. During solo playback, initiation Figure 6. Boxplots showing variation in the rate of answering a mate rates should be high during opposite-sex playback so that to form a duet during both types of solo playback for (a) male and mates can locate each other, and answer rates should be (b) female subjects (FS Z female song, MS Z male song). See Fig. 2 high during same-sex playback. All of these predictions legend for boxplot symbols. were upheld in the present study. Furthermore, an individual’s answer rate should be positively correlated with the risk of misdirected attack. This may explain why song trials. Relative to male song playback, female song the IBD was negatively correlated with answer rate in both playback provoked females to answer at a marginally sexes. Another prediction of the identity hypothesis is higher level (Fig. 6b). Birds of both sexes were more likely that a bird that fails to duet may come into physical to answer their mates’ songs as the distance between them conflict with its own mate during a territorial intrusion. decreased (Fig. 4). Owings & Morton (1998) reported that a captive male Having described how black-bellied wrens duet during Carolina wren, T. ludovicianus, will attack and kill its mate simulated territorial intrusions, we now consider unless the mate answers the male’s song. One of us why they do so. The acoustic mate-guarding hypothesis (D.M.L.) observed a similar incident during a single- (Sonnenschein & Reyer 1983; see Smith 1994 for a varia- speaker duet playback to a pair of free-living black-bellied tion of this hypothesis) has received considerable atten- wrens. Both birds were perched near the playback speaker tion in recent years. The acoustic mate-guarding when the male peered at his mate, flew up, seized her by hypothesis states that song answering serves to counter the bill, and wrestled her to the ground. The female freed the mate-attracting properties of the mate’s song. Levin herself and immediately initiated a duet. After that point, (1996b) suggested that a higher answer rate during same- neither bird directed aggression towards its mate. sex versus opposite-sex playbacks would be evidence of While promoting the identity hypothesis, we have acoustic mate guarding by the answering sex. Although often faced the question ‘why would certain species use we found that females answer at a higher rate during coordinated song for identification when others achieve same-sex playbacks versus opposite-sex playbacks, we the same result using simple calls’? We believe that certain believe that an alternative explanation can explain this ecological conditions and evolutionary trajectories might behaviour as well as other behaviours not explained by favour identification via coordinated song over individu- the acoustic mate-guarding hypothesis. ally distinctive calls. Large IBDs and densely vegetated We propose that birds engage in duets during agonistic territories tend to distort quiet, broadband calls more than encounters to prevent intrapair aggression (Farabaugh loud, whistled songs. As a simple alternative, birds might 1982), an idea we term the ‘identity hypothesis’. This use individually distinctive songs, but among song-learn- hypothesis is similar to the ‘contact maintenance hypoth- ing species (e.g. oscine passerines), this type of identifier esis’, an old idea (Skutch 1960, page 118; Thorpe 1973) could be easily copied by another bird, making it evolu- that has received little attention in the last 20 years. The tionarily unstable. A simple ‘duet code’, unique to each contact maintenance hypothesis states that pairs use duet mated pair, in which initiating song types are linked to 730 ANIMAL BEHAVIOUR, 68, 4

answering song types, would be highly effective in allow- improve our understanding of the behavioural ecology of ing the members of a pair to identify each other reliably. tropical birds in general. Although an intruder might be able to bluff a few answers, the song initiator could use several different song types in sequence to update the probability that any individual is Acknowledgments his mate. The most compelling evidence that certain species duet We thank Gene Morton for his ample assistance through- for mate identification is that a great number of duetting out this study and Mike Baker for all of his valuable advice. species use duet codes rather than combining duet con- Thanks to Michelle Hall, Leah Katz and Jessica Meehan for tributions at random. Buff-breasted wrens, T. leucotis lively discussion and assistance in the field. Mike Antolin, (Farabaugh 1983), bay wrens (Levin 1996a), eastern whip- Bill Black and Shelly Ballard de Volo assisted in the genetic birds, Psophodes olivaceus (Watson 1969), bell shrikes, sexing procedure. D.M.L. received generous support from Laniarius aethiopicus (Thorpe 1973), white-crested laugh- Smithsonian Tropical Research Institute, the Cooper ing thrushes, Garrulax leucolophus (Soucˇek & Vencl 1975), Ornithological Society and an Abbott grant to E. S. African drongos, Dicrurus adsimilis (Wickler 1976), and Morton administered via the Smithsonian Office of slate-coloured boubou shrikes, Laniarius funebris (Wickler Fellowships and Grants. D.E.G. received support from 1976), all show nonrandom song-type associations and/or a National Science Foundation Graduate Research Fellow- pair-specificity in duets. D. M. Logue (unpublished data) ship. Permission to work with birds was granted to D.M.L. confirmed that female black-bellied wrens adhere strictly by the Animal Care and Use Committee of Colorado State to individually unique duet codes. University (Protocol number 01-105A-01) and by Autor- The model that we developed in the Introduction idad Nacional del Ambiente in the Republic of Panama suggests that intersexual territorial defence could arise (Permit number SE/A 034-02). Permission to work at the under realistic conditions. The model made the novel study site was granted by the Panama Canal Railway predictions that males are more likely than females to Company. show intersexual territory defence and that birds engaged in a parenting effort are more likely to defend intersexu- ally than those that are not. Both predictions of the model References were upheld in our study. In accordance with the first prediction, stereo duet playbacks revealed that females, Bard, S. C., Hau, M., Wikelski, M. & Wingfield, J. C. 2002. Vocal but not males, approached the same-sex stimulus more distinctiveness and response to conspecific playback in the spotted closely than the opposite-sex stimulus (Fig. 3). Solo song antbird, a Neotropical suboscine. Condor, 104, 387–394. playbacks revealed a similar tendency, with females Brown, R. N. & Lemon, R. E. 1979. Structure and evolution of song approaching solo female song more closely than male form in the wrens Thryothorus sinaloa and T. felix. Behavioral song, whereas males approached both types of solo song. Ecology and Sociobiology, 5, 111–131. Additionally, during female song playback, males with Charif, R. A., Mitchell, S. & Clark, C. W. 1995. Canary 1.2 User’s dependent juveniles approached the speaker more closely Manual. Ithaca, New York: Cornell Laboratory of Ornithology. than males without juveniles (Fig. 5a). Given that black- Farabaugh, S. M. 1982. The ecological and social significance of duetting. In: Acoustic Communication in Birds Vol. 2. (Ed. by bellied wren fledglings require a great deal of parental care D. E. Kroodsma & E. H. Miller), pp. 85–124. New York: Academic and that replacement females do not aid in caring for the Press. juveniles (D. M. Logue, unpublished data; see Freed 1987 Farabaugh, S. M. 1983. A comparative study of duet song in for a detailed description of mate replacement in house tropical Thryothorus wrens. Ph.D. thesis, University of Maryland. wrens), it is unlikely that this strong male response was an Freed, L. A. 1987. The long-term pair bond of tropical house wrens: attempt to encourage the intruding female to join him on advantage or constraint? American Naturalist, 130, 507–525. the territory. It is much more plausible that fathers Greenberg, R. & Gradwohl, J. 1983. Sexual roles in the dot-winged attempt to oust intruding females because it is in their antwren (Microrhopias quixensis), a tropical forest passerine. Auk, best interest to retain their current mates. We did not find 100, 920–925. any evidence that the presence of juveniles encourages Hall, M. L. 2000. The function of duetting in magpie-larks: conflict, opposite-sex territoriality in females. Furthermore, females cooperation, or commitment? Animal Behaviour, 60, 667–677. did not approach the speaker during male song playback doi:10.1006/anbe.2000.1517. (Fig. 5b), perhaps because size dimorphism in this species Hultsch, H. & Todt, D. 1984. Spatial proximity between allies: renders females incapable of defending territories against a territorial signal tested in the monogamous duet singer Cossypha males without incurring unacceptably high costs. heuglini. Behaviour, 91, 286–293. In conclusion, we wish to emphasize that intrasexual Kahn, N. W., St John, J. & Quinn, T. W. 1998. Chromosome- specific intron size differences in the avian CHD gene provide an territoriality is generally stronger than intersexual territo- efficient method for sex identification in birds. Auk, 115, 1074– riality. The results of our playback experiments, however, 1078. suggest that intersexual territorial defence occurs at non- Levin, R. N. 1996a. Song behaviour and reproductive strategies in trivial levels in this species. Its occurrence in males with a duetting wren, Thryothorus nigricapillus: I. Removal experiments. juveniles is consistent with a model that predicts that Animal Behaviour, 52, 1093–1106. intersexual territory defence can be an optimal strategy in Levin, R. N. 1996b. Song behaviour and reproductive strategies in territorial, socially monogamous species. Understanding a duetting wren, Thryothorus nigricapillus: II. Playback experi- the roles of the sexes in territory defence promises to ments. Animal Behaviour, 52, 1107–1117. LOGUE & GAMMON: VOCAL DUETTING AND TERRITORIAL DEFENCE 731

McGregor, P. K. 1992. Quantifying responses to playback: one, Skutch, A. F. 1960. Life Histories of Central American Birds. Vol. 2. many, or composite multivariate measures? In: Playback and Berkeley, California: Cooper Ornithological Society. Studies of Animal Communication (Ed. by P. K. McGregor), pp. 79– Smith, W. J. 1994. Animal duets: forcing a mate to be attentive. 96. New York: Plenum. Journal of Theoretical Biology, 166, 221–223. Minitab. 2000. Statistics, Release 13.20 for Windows. State College, Sonnenschein, E. & Reyer, H. U. 1983. Mate-guarding and other Pennsylvania: Minitab. functions of antiphonal duets in the slate-coloured boubou Morton, E. S. 1982. Grading, discreteness, redundancy and (Laniarus funebris). Zeitschrift fu¨r Tierpsychologie, 63, 112–140. motivational-structural rules. In: Evolution and Ecology of Acoustic Soucˇek, B. & Vencl, F. 1975. Bird communication study using Communication in Birds (Ed. by D. E. Kroodsma & E. H. Miller), digital computer. Journal of Theoretical Biology, 49, 147–172. pp. 183–212. New York: Academic Press. Stiles, F. G. & Skutch, A. F. 1989. A Guide to the Birds of Costa Rica. Morton, E. S. & Derrickson, K. C. 1996. Song ranging by the dusky Ithaca, New York: Cornell University Press. antbird, Cercomacra tyrannina: ranging without song learning. Stoddard, P. K. 1996. Vocal recognition of neighbors by territorial Behavioral Ecology and Sociobiology, 39, 195–201. passerines. In: Evolution and Ecology of Acoustic Communication in Naguib, M. 1995. Auditory distance assessment of singing Birds (Ed. by D. E. Kroodsma & E. H. Miller), pp. 356–376. Ithaca, conspecifics in Carolina wrens: the role of reverberation and New York: Comstock. frequency-dependent attenuation. Animal Behaviour, 50, 1297– Thorpe, W. H. 1973. Duet-singing birds. Scientific American, 229, 1307. 70–79. Owings, D. H. & Morton, E. S. 1998. Animal Vocal Communication: Watson, M. 1969. Significance of antiphonal song in the eastern a New Approach. Cambridge: Cambridge University Press. whipbird, Psophodes olivaceus. Behaviour, 35, 157–178. Ridgely, R. S. & Gwynne J. A., Jr. 1989. A Guide to the Birds of Wickler, W. 1976. Duetting songs in birds: biological significance of Panama. 2nd edn. Princeton, New Jersey: Princeton University stationary and non-stationary processes. Journal of Theoretical Press. Biology, 61, 493–497. Seddon, N., Butchart, S. H. M. & Odling-Smee, L. 2002. Duetting Wiley, R. H. & Wiley, M. S. 1977. Recognition of neighbors’ duets in the subdesert mesite Monias benschi: evidence for acoustic mate by stripe-backed wrens Campylorhychus nuchalis. Behaviour, 62, defense? Behavioral Ecology and Sociobiology, 52, 7–16. 10–34. doi:10.1007/s00265-002-0488-9.